Time¶
Note
This documentation is mirrored from the graph-scheduler package and often refers to nodes, edges , and graphs. In PsyNeuLink terms, nodes are Mechanisms or Compositions, edges are Projections, and graphs are Compositions. The one exception is during learning, in which Projections may be assigned for execution as nodes to ensure that MappingProjections are updated in the proper order.
Note
This documentation was modified from the original due to environment-specific TimeScale renamings. If there is any confusion, please see the original documentation at https://www.github.com/kmantel/graph-scheduler
Overview¶
Schedulers maintain Clock objects to track time. The current time in relation to a Scheduler is stored in Clock.time or Clock.simple_time
- class psyneulink.core.scheduling.time.TimeScale(value)¶
Represents divisions of time used by the Scheduler and Conditions.
The values of TimeScale are defined as follows (in order of increasingly coarse granularity):
- TIME_STEP¶
the nuclear unit of time, corresponding to the execution of all nodes allowed to execute from a single consideration set of a
Scheduler, and which are considered to have executed simultaneously.
- PASS¶
a full iteration through all of the consideration sets in a
Scheduler'sconsideration_queue, consisting of one or moreTIME_STEPs, over which every node specified to a Scheduler is considered for execution at least once.
- TRIAL¶
an open-ended unit of time consisting of all actions that occurs within the scope of a single call to
run(a single input to a Composition.)
- RUN¶
the scope of a batch of one or more TRIALs. In PsyNeuLink, this is the scope of a call to the
runmethod of a Composition, consisting of one or moreTRIALs.
- LIFE¶
the scope of time since the creation of an object.
- classmethod get_parent(time_scale)¶
- Returns
the TimeScale one level wider in scope than time_scale
- Return type
- class psyneulink.core.scheduling.time.Time(consideration_set_execution=0, pass_=0, environment_state_update=0, environment_sequence=0, life=0, absolute=<Quantity(0, 'millisecond')>, absolute_interval=<Quantity(1, 'millisecond')>, absolute_time_unit_scale=TimeScale.CONSIDERATION_SET_EXECUTION, absolute_enabled=False, **alias_time_values)¶
Represents an instance of time, having values for each
TimeScale- life¶
the
TimeScale.LIFEvalue- Type
int : 0
- environment_sequence¶
the
TimeScale.RUNvalue- Type
int : 0
- environment_state_update¶
the
TimeScale.TRIALvalue- Type
int : 0
- pass_¶
the
TimeScale.PASSvalue- Type
int : 0
- consideration_set_execution¶
the
TimeScale.TIME_STEPvalue- Type
int : 0
- absolute¶
the absolute time value
- Type
pint.Quantity: 0ms
- absolute_interval¶
the interval between units of absolute time
- Type
pint.Quantity: 1ms
- absolute_time_unit_scale¶
the
TimeScalethat corresponds to an interval of absolute time- Type
TimeScale: TimeScale.TIME_STEP
- absolute_enabled¶
whether absolute time is used for this Time object
- Type
bool : False
- _get_by_time_scale(time_scale)¶
- Parameters
time_scale (
TimeScale) –- Returns
this Time’s value of a TimeScale by the TimeScale enum, rather than by attribute
- Return type
int
- _set_by_time_scale(time_scale, value)¶
- Parameters
time_scale (
TimeScale) –
Sets this Time’s value of a time_scale by the TimeScale enum, rather than by attribute
- _increment_by_time_scale(time_scale)¶
Increments the value of time_scale in this Time by one
- _reset_by_time_scale(time_scale)¶
Resets all the times for the time scale scope up to time_scale e.g. _reset_by_time_scale(TimeScale.ENVIRONMENT_STATE_UPDATE) will set the values for TimeScale.PASS and TimeScale.CONSIDERATION_SET_EXECUTION to 0
- class psyneulink.core.scheduling.time.TimeHistoryTree(time_scale=TimeScale.LIFE, max_depth=TimeScale.ENVIRONMENT_STATE_UPDATE, index=0, parent=None, enable_current_time=True)¶
A tree object that stores a history of time that has occurred at various
TimeScales, typically used in conjunction with aClock- time_scale¶
the TimeScale unit this tree/node represents
- Type
- child_time_scale¶
the TimeScale unit for this tree’s children
- Type
- children¶
an ordered list of this tree’s children
- Type
list[
TimeHistoryTree]
- max_depth¶
the finest grain TimeScale that should be created as a subtree Setting this value lower allows for more precise measurements (by default, you cannot query the number of
TimeScale.TIME_STEPs in a certainTimeScale.PASS), but this may use a large amount of memory in large simulations- Type
- index¶
the index this tree has in its parent’s children list
- Type
int
- parent¶
the parent node of this tree, if it exists. None represents no parent (i.e. root node)
- Type
TimeHistoryTree: None
- total_times¶
stores the total number of units of
TimeScales that have occurred over this tree’s scope. Only contains entries forTimeScales of finer grain than time_scale- Type
dict{
TimeScale: int}
- Parameters
enable_current_time (bool : True) – sets this tree to maintain a Time object. If this tree is not a root (i.e. time_scale is
TimeScale.LIFE)
- increment_time(time_scale)¶
Increases this tree’s current_time by one time_scale
- Parameters
time_scale (
TimeScale) – the unit of time to increment
- get_total_times_relative(query_time_scale, base_indices=None)¶
- Parameters
query_time_scale (
TimeScale) – theTimeScaleof units to be returnedbase_indices (dict{
TimeScale: int}) –a dictionary specifying what scope of time query_time_scale is over. e.g.
base_indices = {TimeScale.ENVIRONMENT_SEQUENCE: 1, TimeScale.ENVIRONMENT_STATE_UPDATE: 5}
gives the number of query_time_scales that have occurred in the 5th
ENVIRONMENT_STATE_UPDATEof the 1stENVIRONMENT_SEQUENCE. If an entry for aTimeScaleis not specified but is coarser than query_time_scale, the latest value for that entry will be used
- Returns
the number of units of query_time_scale that have occurred within the scope of time specified by base_indices
- Return type
int